A pair of colorful laser-cut stacked acrylic Raspberry Pi cases with “Moster” (*) heatsinks arrived, with the intent of dressing up the HP 7475A plotters for their next Show-n-Tell:
Unfortunately, the thermal tape on one of the CPU heatsinks was sufficiently wrinkled to prevent good contact with the CPU:
The seller sent a replacement copper slug with tape on one side. Presumably, they glue it to the heatsink with thermal silicone:
Of which, I have none on hand.
So I did what I should have done originally, which was to drop a few bucks on a lifetime supply of thermally conductive heatsink tape, apply it to the bare side of the slug and stick the slug to the heatsink with their tape:
The blue stuff is the separation film, with the tape being white. It doesn’t match the black tape on the other side, but seems gooey enough to work.
Done!
Despite the heatsink hype, ball grid array chips dissipate most of their heat through their pads (and perhaps a central thermal pad) into the PCB, so sticking a heatsink atop the package is largely decorative, along the lines of hotrod ornamentation.
The epoxy packages used in previous Raspberry Pi iterations had better thermal conductivity to their top surface:
Than the more recent metal-top packages, which surely have inert-gas fill under the lid:
Pix cropped after being pilfered from the Official Raspberry Pi site.
Yes, the heatsink does conduct some heat into the air, even if not nearly as much as you might want.
(*) I’m pretty sure “Moster” was a typo in the original eBay listing which took on a life of its own to become something of an unofficial trademark. All of the search results ship from Duluth, Georgia (USA), regardless of the nominal seller; feel free to draw your own conclusions.
The Smell of Molten Projects in the Morning 
I’ve never been able to get a 3B+/3A+ to go anywhere near thermal throttling, and I don’t have heatsinks on any of them. Your application may be different, though.
The same can’t be said for the 4B, though. The tiny temperature-controlled fan I have on my 4 GB 4B (Pimoroni Fan SHIM) will cut in and out even when the system’s idle.
I saw a note recently discussing how thermal throttling doesn’t appear in any of the usual Linux indicators, because it happens in the binary blob code running on the GPU and isn’t communicated to the OS.
The mysterious
vcgencmdseems useful, if awkward, to figure out what’s happening behind the scenes, although I can never remember the syntax.vcgencmd get_throttledwill show throttling status, though its definition of thermal throttling is a little more loose than most people would like. The Fan SHIM’s daemon just watches the CPU temperature reported in /proc and turns on at a very safe 60°CI meant to add: this is a short guide to the get_throttled values from a Raspberry Pi engineer: https://www.raspberrypi.org/forums/viewtopic.php?f=63&t=147781&start=50#p972790
That looks familiar! [mutter]
The “throttled” value is latched to means “has ever been throttled since boot”. That’s entirely reasonable, as it should never throttle in normal operation.
A quick check of my always-on RPis says they’re all good: thanks for the pointer!